arson 0.0.7

A restructured object notation

ARSON: A Restructured Object Notation

ARSON is JSON, with a little bit of sugar: Comments, Commas, and Tags.

For example:

{
    "numbers": +0123.0,       # Can have leading zeros
    "octal": 0o10,            # Oh, and comments too
    "hex": 0xFF,              #
    "binary": 0b1000_0001,    # Number literals can have _'s 

    "lists": [1,2,3,],        # Lists can have trailing commas

    "strings": "At least \x61 \u0061 and \U00000061 work now",
    "or": 'a string',         # both "" and '' work.

    "records": {
        "a": 1,               # Must have unique keys
        "b": 2,               # and the order must be kept
    },
}

Along with some sugar atop JSON, ARSON supports tagging literals to represent types outside of JSON:

• @datetime "2017-11-22T23:32:07.100497Z", a tagged RFC 3339 datestamp
• @duration 60 (a duration in seconds, float or int)
• @base64 "...==", a base64 encoded bytestring
• @set, @dict, @complex, @bytestring

Quickstart

Use pip install arson to get the latest copy of the python library. The arson library has two methods, parse and dump.

import arson

print(arson.parse(arson.dump([1,2,3])))

If you want to use your own tagged object types, you can create a custom Codec:

import arson

class Example:
    def __init__(self, value):
        self.value = value
    def __repr__(self):
        return f"Example {self.value}"

def object_to_tagged(obj):
    if isinstance(obj, Example):
        return "Example", {"value": obj.value}
    else:
        raise NotImplementedError()

def tagged_to_object(name, value):
    if name == "Example":
        return Example(value['value'])
    else:
        raise NotImplementedError()
        
codec = arson.Codec(object_to_tagged, tagged_to_object)

print(codec.parse(codec.dump(Example(1))))

Supported Datatypes

This library supports serializing and deserializing the following types

• str, int, bool
• float (including Infinity and Nan)
• list, tuple (but will always return a list)
• dict, set, OrderedDict (but will use dicts on 3.7+)
• datetime, timedelta (but will convert times to UTC)
• bytestring, bytearray (but will return a bytearray)
• complex

The fixed width numeric types are currently unsupported (Sorry).

ARSON in a Nutshell

• File MUST be utf-8, not cesu-8/utf-16/utf-32, without surrogate pairs.
• Use #.... <end of line> for comments
• Byte Order Mark is treated as whitespace (along with \x09, \x0a, \x0d, \x20)
• ARSON Document is any ARSON Object, (i.e 1 is a valid ARSON file).
• Lists are [], [obj], [obj,], [obj, obj] ... (trailing comma optional)
• Records are {"key": value}, keys must be unique, order must be preserved.
• Built-ins: true, false, null
• "unicode strings" with escapes \" \\ \/ \b \f \n \r \t \uFFFF \UFFFFFFFF, no control codes unecaped, and '' can be used instead of "".
• int/float numbers (unary plus or minus, allowleading zeros, hex, octal, and binary integer liters)
• Tagged literals: @name [1,2,3] for any other type of value.

Errors are fatal. A record with duplicate keys, or a string too long, or a number too big to represent MUST cause the parse to fail outright.